Kévin Petit

Did your build complete successfully? It seems some tests were built but not others.

Thanks for the feedback and suggestions! It's good to know that CI artefacts are good enough for some people at least. I'll look into publishing artefacts from the existing Azure...

I've now pushed a basic change to the CI to publish artifacts. This will be done for each commit to the master branch. You can access them by clicking on...

Useful report, good to know, thanks! You may want to run with logging enabled (passing `CLVK_LOG=4` in the environment) to try and get more info about what's failing.

Thanks for the feedback! 1. Yes, my prototype has them disabled entirely but this only worked because I was looking at a very narrow set of workloads. 2. Vulkan only...

My experience is that even with recent versions, it does occasionally happen. I've found that removing the `errs() ...` before the `llvm_unreachable` seems to suppress the infinite loop. Linking #383...

@alan-baker This will be needed by clvk ultimately to support `clGetKernelArgInfo` (see https://www.khronos.org/registry/OpenCL/sdk/2.1/docs/man/xhtml/clGetKernelArgInfo.html). There is a Clang option to emit all the information required as metadata (see `EmitOpenCLArgMetadata` in `Compiler.cpp`)....

I'm not sure I'm following. Unless I'm missing something, the issue I'm seeing is that libclc is not passing conformance but the glsl extended instruction is. Looking at the implementation...

I see, makes sense. Thank you for the explanation. I'm amazed I'm only bumping into this now! I'll push a fix to clvk.

Thanks for the report. The short answer is that clspv doesn't support the vload/vstore built-in functions in C++ code unless the pointer is in the constant address space. For the...